1998 Volume 41 Issue 4 Pages 840-846
Active control of jets emitted into a two-dimensional channel with insulated suction near the wall during monitoring with a neural network is described. The aim of this work is to simulate the possibility of controlling the flow pattern with a flip-flop phenomenon by suitable suction. In this case, the output of a neural network is estimated satisfactorily based on 3 flow patterns of teaching data. The periodic flow pattern at the Reynolds number Re=1.0×103 is reduced by a suction rate of less than 10% for primary jets. Then flow becomes stable as a result of feedback gain of the neural network. Since the control with suction need not be continuous, the power cost is reduced. The nonperiodic case at Re=1.0×104 is difficult to control, but we achieved good control by with the use of a suitable suction rate with the neural network.